Document 10552486

advertisement
Journal of Lie Theory
Volume 12 (2002) 551–570
c 2002 Heldermann Verlag
Moduli for Spherical Maps and Minimal Immersions
of Homogeneous Spaces
Gabor Toth
Communicated by F. Knop
Abstract.
The DoCarmo-Wallach theory studies isometric minimal immersions f : G/K → S n of a compact Riemannian homogeneous space G/K into
Euclidean n -spheres for various n . For a given domain G/K , the moduli space
of such immersions is a compact convex body in a representation space for the
Lie group G . In 1971 DoCarmo and Wallach gave a lower bound for the (dimension of the) moduli for G/K = S m , and conjectured that the lower bound was
achieved. In 1997 the author proved that this was true. The DoCarmo-Wallach
conjecture has a natural generalization to all compact Riemannian homogeneous
domains G/K . The purpose of the present paper is to show that for G/K a
nonspherical compact rank 1 symmetric space this generalized conjecture is false.
The main technical tool is to consider spherical functions of subrepresentations
of C ∞ (G/K), express them in terms of Jacobi polynomials, and use a recent
linearization formula for products of Jacobi polynomials.
1.
Introduction and Statement of Results
Let M = G/K be a Riemannian homogeneous space, where G is a compact Lie
group and K a closed subgroup. Then G acts on the space C ∞ (M ) of (real
valued) functions on M in a natural way: g · ξ = ξ ◦ g −1 , g ∈ G, ξ ∈ C ∞ (M ).
This action preserves the L2 -scalar product on C ∞ (M ) defined by the volume
element vM . Let H ⊂ C ∞ (M ) be a G-submodule. We call a map f : M → SV
into the unit sphere SV of a Euclidean vector space V a spherical H-map if its
components α ◦ f , α ∈ V ∗ , belong to H. The Dirac delta δ : M → SH∗ [5]
defined by evaluating the elements of H on points of M is the universal example
of a spherical H∗ -map. (The scalar product on H∗ is induced by the L2 -scalar
product on H suitably scaled.)
Remark.
If M = G/K is naturally reductive, and H ⊂ C ∞ (M ) is irreducible
then H is contained in an eigenspace Vλ of the Laplacian 4M for some eigenvalue
λ [25]. In particular, the components of an H-map f : M → SV are eigenfunctions
of the Laplacian with a common eigenvalue. Thus an H-map is a λ-eigenmap in
the sense of Eells-Sampson [8], a harmonic map with constant energy density λ/2.
ISSN 0949–5932 / $2.50
c Heldermann Verlag
552
Toth
In general, a DoCarmo-Wallach type argument [6] shows that the set of (congruence classes of) full spherical H-maps f : M → SV , for various V , can be
parametrized by a moduli space L(H), a compact convex body in a G-submodule
E(H) of the symmetric square S 2 (H) (Propositions 3.1-3.2 in Section 3 below).
(The map f is full if its image is not contained in a proper great sphere of the
range [3], and congruent maps differ by an isometry between the ranges.) In what
follows, we identify S 2 (H) with the space of linear endomorphisms of H. Then
the moduli space is given by
L(H) = {C ∈ E(H) | C + I ≥ 0},
where ≥ means positive semidefinite, and I is the identity. The origin of E(H) is
in the interior of L(H), and it corresponds to δ .
The G-module homomorphism
Ψ0 : S 2 (H) → C ∞ (M )
given by multiplication has image H · H ⊂ C ∞ (M ) consisting of (finite) sums
of products of functions in H. The DoCarmo-Wallach parametrization of L(H)
implies that the kernel of Ψ0 is E(H) (Proposition 3.3). We thus have
E(H) = S 2 (H)/(H · H),
as G-modules. To determine E(H) (and thereby to compute dim L(H) = dim E(H))
amounts to decomposing H · H into irreducible components.
Let M = G/K be a compact rank 1 symmetric space. Then M is the Euclidean
m-sphere S m , one of the projective spaces RP m , CP m , HP m , or the Cayley
projective plane CaP 2 [2]. It is well-known that C ∞ (M ) has a multiplicity one
decomposition into irreducible components, and each component H ⊂ C ∞ (M ) is
the full eigenspace Vλ of the Laplacian corresponding to an eigenvalue λ [15-17].
Our first result is the following:
Theorem.
A Let M = G/K be a compact rank 1 symmetric space, H ⊂
C ∞ (M ) an irreducible G-submodule. We write H = Vλp , where λp is the p-th
eigenvalue of the Laplacian on H. Then we have
Pp
Vλ2j if M = S m
Vλp · Vλp = Pj=0
2p
otherwise.
j=0 Vλj
In particular, the dimension of the moduli space is given by
P
n(λp )(n(λp ) + 1)/2 − pj=0 n(λ2j ) if M = S m ,
P
dim L(Vλp ) =
n(λp )(n(λp ) + 1)/2 − 2p
otherwise,
j=0 n(λj )
where n(λp ) = dim Vλp .
Since n(λp ) is known for each case of M (second table in Section 2), an explicit
formula can be derived for the dimension of L(Vλp ). If the dimension is zero then
the moduli space reduces to a point, and we have rigidity. This means that the
corresponding spherical Vλp -maps are rigid in the sense that any full f is congruent to the Dirac delta δ .
553
Toth
Corollary.
Let M and H = Vλp be as
dim L(Vλp ) is trivial are summarized in the
M
Sm
m
S , RP m
CP m , HP m , CaP 2
in Theorem A. Then the cases when
following table:
m
p
m≥2 p=1
m=2 p≥1
m=2 p=1
Remark.
Rigidity of a spherical Vλ1 -map f : S m → SV is obvious since f is the
restriction of a linear map, and thereby it is an isometry. Rigidity of spherical Vλp maps f : M → SV for M = S 2 , RP 2 is due to Calabi [3] (stated only for minimal
immersions). (In general, a spherical Vλp -map f : RP m → SV is a spherical Vλ2p map f˜: S m → SV factored through the twofold projection S m → RP m .)
A rigidity result of DoCarmo-Wallach [6,25] asserts that a minimal immersion
f : M → SV of a compact analytic manifold M is rigid among minimal immersions,
if the (geometric) degree of f is < 4. For M = CP m , HP m , CaP 2 as in the
corollary, the degree of δ : M → SVλp is 2p [18]. Notice however that the corollary
gives rigidity among all spherical Vλ1 -maps not just minimal immersions.
We now return to the general setting. Let G be a compact Lie group. An
orthogonal G-module H is a Euclidean vector space on which G acts linearly
via orthogonal transformations. In other words, H is a representation space for
G, and it is endowed with a G-invariant scalar product.
Let K be a closed subgroup. A class 1 representation of (G, K) is an irreducible
orthogonal G-module H so that there is a nonzero vector χ0 ∈ H fixed by K .
It is well known that, for M = G/K Riemannian homogeneous, the irreducible
components of C ∞ (M ) are class 1 representations of (G, K).
Since all components of C ∞ (M ) are class 1 with respect to (G, K) it is natural
to ask whether H · H ⊂ C ∞ (M ) contains all class 1 components of S 2 (H).
We reformulate this by introducing Ē(H) as the sum of those irreducible Gsubmodules of S 2 (H) that are not class 1 with respect to (G, K). The very
existence of the homomorphism Ψ0 above implies that
Ē(H) ⊂ E(H),
and the question is whether equality holds. For M = S m , the answer is yes, and it
follows from the (multiplicity one) decomposition for S 2 (Vλp ) derived by DoCarmo
and Wallach [6]. (For a simple proof, see also [14].)
Our next result shows that the answer is negative for M = CP m .
Theorem.
B Let (G, K) = (U (m + 1), U (m) × U (1)), and M = CP m , m ≥ 2,
the complex projective m-space. Let H = Vλp , p ≥ 2. Then E(Vλp ) contains class
1 submodules with respect to (U (m + 1), U (m) × U (1)). Equivalently
Ē(Vλp ) 6= E(Vλp ).
More precisely, we have
2p−2
X 1
(−1)q − 1
min (q, 2p − q) +
Vλq ⊂ E(Vλp ).
2
2
q=2
554
Toth
Remark.
For M = CP m , Theorems A-B correct Theorem 4.2 of Chapter III
in [22]. The formula there should give only a lower bound for E(Vλp ) and for the
dimension of the moduli space L(Vλp ). This is due to the fact that the osculating
spaces of δ : CP m → SVλ∗p are reducible as U (m)-modules (Theorem 3.3 of Chapter
II).
Assume now that M = G/K is isotropy irreducible. This means that K acts on
the tangent space To (M ), o = {K}, irreducibly by the isotropy representation.
Then, for an irreducible G-submodule H ⊂ C ∞ (M ), the Dirac delta δ : M → SH∗
is a minimal immersion inducing the λ/ dim M -multiple of the original Riemannian
metric on M [25].
DoCarmo and Wallach proved that the set of (congruence classes of) full minimal
immersions f : M → SV , for various V , and with induced Riemannian metric the
λ/ dim M -multiple of the original, can be parametrized by a moduli space M(H),
a compact convex body in a G-submodule F(H) of S 2 (H) (Proposition 3.4). The
moduli space is given by
M(H) = {C ∈ F(H) | C + I ≥ 0},
where ≥ means positive semidefinite.
We now recall the definition of induced representations [25]. If W is a K -module
then IndG
K (W) denotes the linear space of continuous maps φ : G → W which
satisfy φ(kg) = k · φ(g), k ∈ K , g ∈ G. The action of G on IndG
K (W) given by
0
0
0
g · φ(g ) = φ(gg ), g, g ∈ G, defines a G-module structure on IndG
K (W). We call
G
IndK (W) the G-module induced from the K -module W .
DoCarmo and Wallach constructed a homomorphism
2
Ψ : S 2 (H) → IndG
K (S (p)),
where K -module S 2 (p) is the symmetric square of the isotropy representation of
M = G/K . The kernel of Ψ is F(H).
Let F̄(H) denote the sum of those components of S 2 (H) that, when restricted to
K , do not contain any irreducible K -submodules of S 2 (p). Frobenius reciprocity
[25] says that
F̄(H) ⊂ F(H).
(1)
Thus, once the irreducible decomposition of S 2 (H) is known, this gives a lower
bound on the dimension of the moduli M(H).
DoCarmo and Wallach carried this out for M = S m , and H = Vλp . Identifying
the irreducible components of F̄(Vλp ), for m ≥ 3 and p ≥ 4, they obtained the
lower estimate
dim M(Vλp ) = dim F(Vλp ) ≥ dim F̄(Vλp ) ≥ dim F̄(Vλ4 ) ≥ 18.
They conjectured that equality holds in (1). This has been resolved by the author
in [21] using different methods. (For a recent algebraic proof, see [26].) For the
lowest dimensional moduli space, M(Vλ4 ) with m = 3, see [23].
Once again, it is natural to ask whether equality holds in (1) in general, or at least
for compact rank 1 symmetric spaces M = G/K . Our last result is to show that
the answer is negative for M = CP m , and H = Vλp .
555
Toth
Theorem.
C Let m ≥ 3, (G, K) = (U (m + 1), U (m) × U (1)), M = U (m +
1)/(U (m) × U (1)) = CP m , and H = Vλp . Then, for p = 3 and m 6≡ 1 (mod 4),
or for p ≥ 4, we have
F̄(Vλp ) 6= F(Vλp ).
The striking difference between the spherical and complex projective cases is that
S 2 (Vλp ), Vλp ⊂ C ∞ (S m ), has a multiplicity one decomposition into irreducible
components, but according to the multiplicity formulas developed by Barbasch
[22], this fails for S 2 (Vλp ), Vλp ⊂ C ∞ (CP m ).
2.
Zonal Spherical Functions and Jacobi Polynomials
In this section we describe the main idea of the proof of Theorem A as well as
assemble some preliminary facts.
Let M = G/K be a compact rank 1 symmetric space. As noted above, an
irreducible G-submodule H ⊂ C ∞ (M ) is class 1 with respect to the pair (G, K).
We call a K -fixed vector χ0 ∈ H a zonal spherical function [15,25]. It is wellknown that a zonal spherical function is unique up to a constant multiple [2].
Let χ0 be a zonal spherical function of H. Its square χ20 ∈ H·H is also fixed by K .
Since C ∞ (M ) has a multiplicity one decomposition into irreducible components,
as an element of C ∞ (M ), χ20 decomposes into a sum
χ20
=
n
X
χj ,
j=1
where each χj belongs to a unique irreducible component Hj ⊂ C ∞ (M ). Clearly,
χj is a zonal spherical function of Hj . Since χj ∈ Hj is a component of χ20 ∈ H·H,
by Schur’s lemma, Hj projects nontrivially to H · H, and we obtain
n
X
Hj ⊂ H · H.
j=1
In Section 4 we will prove Theorem A by showing that equality holds here.
We now illustrate this in a different setting by a simple example.
Example.
Let G be a compact Lie group viewed as a symmetric space G ×
∗
G/G of compact type, where G∗ ⊂ G × G is the diagonal [15,16,24]. (The
map (g1 , g2 )G∗ 7→ g1 g2−1 , g1 , g2 ∈ G, identifies G × G/G∗ with G.) The space
C ∞ (G × G/G∗ , C) of complex valued smooth functions on G × G/G∗ has a
multiplicity one decomposition into irreducible components. A component, a
complex irreducible G×G-submodule of C ∞ (G×G/G∗ , C), has the form H∗ ⊗H,
where H is a complex irreducible G-module. The G∗ -fixed vectors in H∗ ⊗ H can
be identified with the (multiples of a normalized) character χ0 of H [24]. Given
χ0 , according to our procedure, we need to decompose the square χ20 into a sum
of (nonzero) characters
n
X
2
χ0 =
cj χj .
j=1
556
Toth
By elementary character theory, this decomposition corresponds to the decomposition of the tensor product
H⊗H=
n
X
cj Hj
j=1
as a G-module, where χj is the character of Hj and cj ∈ N is the multiplicity of
Hj in H ⊗ H. Since χ20 ∈ (H∗ ⊗ H) · (H∗ ⊗ H), Schur’s lemma tells us that
n
X
Hj∗ ⊗ Hj ⊂ (H∗ ⊗ H) · (H∗ ⊗ H)
j=1
as G×G-modules. We claim that equality holds here. Indeed, consider the natural
extension of Ψ0 above
Ψ0 : (H∗ ⊗ H) ⊗ (H∗ ⊗ H) → (H∗ ⊗ H) · (H∗ ⊗ H)
given by multiplication. The domain of Ψ0 , as a G×G-module, can be decomposed
as
(H∗ ⊗ H) ⊗ (H∗ ⊗ H) = (H∗ ⊗ H∗ ) ⊗ (H ⊗ H)
!
!
n
n
n
X
X
X
∗
=
cj Hj ⊗
cl Hl =
cj cl Hj∗ ⊗ Hl .
j=1
l=1
j,l=1
Finally, by Schur’s lemma again Hj∗ ⊗ Hl contains a G∗ -fixed vector if and only if
j = l [24]. The claim follows.
We now return to our compact rank 1 symmetric space M = G/K . As noted in
Section 1, the full eigenspace Vλ of the Laplacian 4M corresponding to an eigenvalue λ is an irreducible G-module. Moreover, if {λp }∞
p=0 denotes the sequence of
eigenvalues of 4M in increasing order, then we have
∞
C (M ) =
∞
X
Vλp .
p=0
By the above, Vλp ⊂ C ∞ (M ) contains a zonal spherical function χ0 , unique up to
a constant multiple.
(α,β)
We now recall that, for fixed α, β > −1, the Jacobi polynomials Pn , n ≥ 0,
form an orthogonal series on [−1, 1] with respect to the weight function (1 −
(α,β)
x)α (1 − x)β [1]. The polynomial Pn
can be defined by
(1 − x)α (1 − x)β Pn(α,β) (x) =
(−1)n dn
[(1 − x)n+α (1 − x)n+β ].
2n n! dxn
With a suitable choice of parameters on M , the zonal function χ0 of a component
(α,β)
Vλp of C ∞ (M ) is a constant multiple of Pn
with α, β , n depending on Vλp
[4,10,24]. (For example, n = p in all cases but M = RP m for which n = 2p.)
The classification of compact rank 1 symmetric spaces M , the eigenvalues of 4M
[2], and the Jacobi polynomials corresponding to the zonal spherical harmonics χ0
are summarized in the following table:
557
Toth
(G, K)
M = G/K
λp
χ0
(m/2−1,m/2−1)
m
(SO(m + 1), SO(m))
S
p(p + m − 1) Pp
(m/2−1,m/2−1)
m
(SO(m + 1), O(m))
RP
2p(2p + m − 1) P2p
(m−1,0)
(U (m + 1), U (m) × U (1))
CP m
4p(p + m)
Pp
(2m−1,1)
(Sp(n + 1), (Sp(n) × Sp(1))
HP m
4p(p + 2m + 1)
Pp
(7,3)
(F4 , Spin(9))
CaP 2
4p(p + 11)
Pp
The multiplicities n(λp ) = dim Vλp are given as follows:
M
)
n(λpp+m−2
p+m
Sm
−
m m
2p+m
RP m
− 2p+m−2
m
m 2
p+m 2
CP m
− p+m−1
m
m
2p+2m+1 p+2m p+2m−1
HP m 2m(2m+1)
2m−1
2m−1
2p+11 p+10 p+7
CaP 2
1320
7
7
To simplify the treatment and to avoid some overlapping cases, we will assume
that m ≥ 2.
Since, up to parametrization, the zonals are Jacobi polynomials, we need to obtain
(α,β)
a decomposition of the square (Pp )2 into a sum of Jacobi polynomials:
(Pp(α,β) )2
=
2p
X
(α,β)
c(j, p; α, β)Pj
.
(2)
j=0
More generally, a formula of the type
Pp(α,β) Pq(α,β) =
p+q
X
(α,β)
c(j, p, q; α, β)Pj
.
(3)
j=|p−q|
is usually called “linearization of the product.”
(α,β)
For α = β , the Jacobi polynomial Pp
is, up to normalization, the ultraspherical
ν
(or Gegenbauer) polynomial Cp , where ν − 1/2 = α = β . (The precise formula
is given in (20) below.) Linearization of the product of ultraspherical polynomials
dates back to the early twentieth century, and the coefficients c(j, p, q; λ − 1/2, λ −
1/2) have been calculated explicitly [1,7,24]. For our purposes, we need only that
c(j, p, q; λ − 1/2, λ − 1/2) is positive if and only if |p − q| ≤ j ≤ p + q and j ≡ p + q
(mod 2).
For Jacobi polynomials in general linearization proved to be much more difficult
and the exact decomposition formula is fairly recent [19]. A general and sharp
positivity result for the coefficients c(j, p, q; α, β) (covering the remaining cases in
the table above for m ≥ 2) is due to Gasper [11,12]. It states that if α, β > −1,
a = α + β + 1, b = α − β , then c(j, p, q; α, β) > 0 provided that (α, β) is in the
interior of the set
V = {(α, β) | α ≥ β, a(a + 5)(a + 3)2 ≥ (a2 − 7a − 24)b2 }.
Note that Theorem 1 in [12] states nonnegativity of the coefficients for (α, β) ∈ V .
As Professor Gasper communicated to the author [13], a closer inspection of his
proof of Theorem 1 in [12], pp. 585-591, shows strict positivity of the coefficients
if (α, β) is in the interior of V . Another proof of the positivity follows by using
the { }9 F8 series representations for the linearization coefficients in [19] (formula
(3.9)).
558
Toth
3.
Generalities on the Moduli
Let G be a compact Lie group and H an orthogonal G-module. We define
K(H) = {C ∈ S 2 (H) | C + I ≥ 0}.
We write K = K(H) if there is no danger of confusion. K is a G-invariant set in
S 2 (H), where the G-module structure on S 2 (H) is extended from that of H.
Since C + I ≥ 0 is a convex condition, K is a convex set. The interior of K
consists of those endomorphisms C that satisfy C + I > 0. It follows that K
has a nonempty interior, and hence it is a convex body in S 2 (H). Notice that
K is noncompact since the multiples λI , λ ≥ −1, are contained in K. We call
K = K(H) the general moduli space for H.
We let S02 (H) denote the G-submodule of S 2 (H) comprised of the traceless
symmetric endomorphisms of V . We define
K0 = K0 (H) = K(H) ∩ S02 (H) = {C ∈ S02 (H) | C + I ≥ 0}.
The eigenvalues of the symmetric endomorphisms in K are greater or equal to −1.
Hence the eigenvalues of the endomorphisms in K0 are contained in [−1, dim H−1].
It follows that K0 is compact, and a convex body in S02 (H). We call K0 = K0 (H)
the reduced moduli space for H.
We now give an interpretation of the moduli as parameter spaces for certain maps.
We let M be a compact Riemannian manifold, and G a compact Lie group of
isometries of M . (G is a closed subgroup of the full isometry group of M .) The
space C ∞ (M ) of smooth functions on M is a representation space for G, where
g ∈ G acts on ξ ∈ C ∞ (M ) by g · ξ = ξ ◦ g −1 . We fix a finite dimensional
G-submodule H ⊂ C ∞ (M ). We endow H with the scaled L2 -scalar product
Z
dim H
χ1 χ2 vM , χ1 , χ2 ∈ H,
(4)
hχ1 , χ2 i =
vol (M ) M
R
where vM is the Riemannian volume form on M , and vol (M ) = M vM is the
volume of M . With this scalar product H becomes an orthogonal G-module.
A smooth map f : M → V into a Euclidean vector space V is said to be full if
the image of f spans V . A component of f is α ◦ f ∈ C ∞ (M ), where α ∈ V ∗ .
The space of components of f is defined as
Vf = {α ◦ f | α ∈ V ∗ } ⊂ C ∞ (M ).
The map f is full if and only if the linear map f ∗ : V ∗ → Vf , given by precomposition with f , is an isomorphism. Since V is Euclidean we also have V ∼
= V∗ ∼
= Vf .
Note that any map can be made full by restricting its range to the linear span of
the image.
Two maps f1 : M → V1 and f2 : M → V2 are said to be congruent if there is a
linear isometry U : V1 → V2 such that f2 = U ◦ f1 .
With H as above, f : M → V is said to be an H-map if Vf ⊂ H. Note that
any smooth map f : M → V is an H-map for H the smallest G-invariant linear
subspace in C ∞ (M ) that contains Vf .
The Dirac delta as a map δH : M → H∗ is defined in the usual way
δH (x)(χ) = χ(x),
x ∈ M, χ ∈ H.
559
Toth
The component of δH corresponding to χ ∈ H = H∗∗ is hδH , χi = χ. Hence,
VδH = H and δH is a full H-map.
In what follows we will identify H with its dual H∗ via the scalar product on H.
With respect to an orthonormal basis {χj }N
j=0 ⊂ H, dim H = N + 1, the Dirac
delta as a map δH : M → H can be written as
δH (x) =
N
X
χj (x)χj ,
x ∈ M.
(5)
j=0
Indeed, for χ ∈ H, we have
hδH (x), χi = χ(x) =
N
X
hχ, χj iχj (x) =
j=0
* N
X
+
χj (x)χj , χ .
j=0
The Dirac delta δH is equivariant with respect to the homomorphism ρH : G →
O(H) that defines the orthogonal G-module structure on H ∼
= H∗ .
For a full H-map f : M → V , we have f = A ◦ δH , where A : H → V is a
surjective linear map. We associate to f the symmetric linear endomorphism
hf i = A∗ A − I ∈ S 2 (H).
It is clear that hf i depends only on the congruence class of f . Since A∗ A is
always positive semidefinite, we also have hf i ∈ K(H). A DoCarmo-Wallach type
argument shows that f 7→ hf i gives rise to a one-to-one correspondence between
the set of congruence classes of full H-maps and the general moduli space K(H)
[6,25].
Let f : M → V be a full H-map. With respect to an orthonormal basis in V ,
f can be written in components as f = (f 0 , . . . , f n ), dim V = n + 1. With the
orthonormal basis {χj }N
j=0 ⊂ H as above, A : H → V becomes an (n+1)×(N +1)matrix with entries akj , k = 0, . . . , n, j = 0, . . . , N . In components, f = A ◦ δH
can be written as
N
X
k
f =
akj χj , k = 0, . . . , n.
j=0
We now calculate
∗
trace (hf i + I) = trace A A =
n X
N
X
a2kj
k=0 j=0
=
n
X
|f k |2 .
k=0
We conclude that, in terms of the scaled L2 -scalar product (4) on H, the parameter
point hf i ∈ K(H) is traceless if and only if
Z X
n
(f k )2 vM = vol (M ).
(6)
M k=0
We call f normalized if (6) is satisfied. Clearly, δH is normalized.
It is also clear that, by suitable scaling, any nontrivial map can be normalized.
Summarizing, we obtain the following:
560
Toth
Proposition 3.1.
Let M be a compact Riemannian manifold with compact
group G of isometries. Given a finite dimensional G-submodule H of C ∞ (M ),
the set of congruence classes of full H-maps f : M → V can be parametrized
by the general moduli space K(H). The reduced moduli K0 (H) parametrizes the
normalized H-maps.
An H-map f : M → V is called spherical if the image of f is contained in the
unit sphere SV of V . A finite dimensional G-module H ⊂ C ∞ (M ) is called δ spherical if δH is spherical. Due to the scaling of the L2 -scalar product in (4), H
is δ -spherical if and only if
N
X
(χj )2 = 1
j=0
on M , where {χj }N
j=0 ⊂ H is an orthonormal basis.
If M = G/K is homogeneous then any H ⊂ C ∞ (M ) is δ -spherical. This is
because δH is equivariant, and thereby its image is a G-orbit in H necessarily
contained in SH .
Let H be a δ -spherical G-module. A full H-map f : M → V is spherical if and
only if
|f (x)|2 − |δH (x)|2 = h(A∗ A − I)δH (x), δH (x)i = hhf i, δH (x)
for all x ∈ M . Here
δH (x)i = 0,
denotes the symmetric tensor product. We define
E(H) = {δH (x)
δH (x) | x ∈ M }⊥ ⊂ S 2 (H).
(7)
The previous computation shows that an H-map f : M → V is spherical if and
only if hf i ∈ E(H).
Once again, since δH is equivariant, E(H) ⊂ S 2 (H) is a G-submodule.
We obtain the following:
Proposition 3.2.
Let M be a compact Riemannian manifold with compact
group G of isometries, and H ⊂ C ∞ (M ) a δ -spherical G-submodule. Then the set
of congruence classes of full spherical H-maps f : M → SV can be parametrized
by the moduli space
L(H) = K(H) ∩ E(H).
Moreover L(H) is a compact convex body in E(H).
Compactness follows since spherical maps are automatically normalized:
L(H) ⊂ K0 (H) ⇒ E(H) ⊂ S02 (H),
so that
L(H) = K0 (H) ∩ E(H).
Remark.
Let M = G/K be a compact naturally reductive Riemannian homogeneous space, and Vλ ⊂ C ∞ (M ) the eigenspace of 4M corresponding to an
eigenvalue λ. Recall from Section 1 that a λ-eigenmap f : M → SV is a spherical
Vλ -map.
561
Toth
Let H ⊂ C ∞ (M ) be a finite dimensional G-submodule. Then H ⊂ Vλ for some
λ. Proposition 3.2 asserts that L(H) parametrizes the congruence classes of full
λ-eigenmaps f : M → SV with components in H ⊂ Vλ . In particular, L(Vλ )
parametrizes the congruence classes of all full λ-eigenmaps f : M → SV .
Returning to the general situation, let H ⊂ C ∞ (M ) be a δ -spherical G-module.
We define
Ψ0 = Ψ0H : S 2 (H) → C ∞ (M )
(8)
by
Ψ0 (C)(x) = hCδH (x), δH (x)i = hC, δH (x)
δH (x)i,
x ∈ M.
Since δH is equivariant, Ψ0 is a homomorphism of G-modules. By (7), we have
ker Ψ0 = E(H).
(9)
We claim that the image of Ψ0 is the G-submodule
H · H = span {χ1 χ2 | χ1 , χ2 ∈ H} ⊂ C ∞ (M ).
Indeed, using (5) in the definition of Ψ0 , we obtain
Ψ0 (C) =
N
X
cjl χj χl ,
j,l=0
where {χj }N
j=0 ⊂ H is an orthonormal basis, and the cjl ’s are the matrix entries
2
of C ∈ S (H). The claim follows.
Note that H · H always contains the trivial G-module, a consequence of δ sphericality.
We obtain the following:
Proposition 3.3.
Let H ⊂ C ∞ (M ) be a δ -spherical G-module. Then the Gmodule homomorphism
Ψ0 : S 2 (H) → H · H
is onto, and has kernel E(H). In particular, H · H is (isomorphic to) a Gsubmodule of S 2 (H) and we have
E(H) ∼
= S 2 (H)/(H · H)
as G-modules.
Let K ⊂ G be a closed subgroup. Recall that an irreducible orthogonal G-module
V is called class 1 with respect to the pair (G, K) if V contains a nonzero K -fixed
vector, or equivalently, if V|K contains the trivial representation.
We now assume that M = G/K is Riemannian homogeneous. As noted in Section
1, any irreducible G-submodule of C ∞ (M ) is class 1 with respect to (G, K).
Conversely, any class 1 G-module V with respect to (G, K) is isomorphic to an
irreducible G-submodule of C ∞ (M ) [25].
Let H ⊂ C ∞ (M ) be a δ -spherical G-submodule. We define Ē(H) ⊂ S 2 (H) as the
562
Toth
sum of those irreducible G-submodules in S 2 (H) that are not class 1 with respect
to (G, K). By the description of class 1 modules above and (8)-(9), we see that
Ē(H) ⊂ E(H).
(10)
Equality holds if and only if the sum of all irreducible G-submodules in S 2 (H)
that are class 1 with respect to (G, K), is isomorphic to H · H.
A map f : M → V is said to be conformal if
hf∗ (X), f∗ (Y )i = chX, Y i,
X, Y ∈ T (M ),
where c > 0 is a constant. Then c is called the conformality constant of f .
We say that a finite dimensional G-module H ⊂ C ∞ (M ) is δ -conformal if δH is
conformal.
Using (5), we have
(δH )∗ (X) = XδH =
N
X
X(χj )χj ,
X ∈ T (M ).
(11)
j=1
Thus, H is δ -conformal if and only if
N
X
X(χj )Y (χj ) = chX, Y i,
X, Y ∈ T (M ),
(12)
j=0
holds for any orthonormal basis {χj }N
j=0 ⊂ H.
Let f : M → V be a conformal map as above, and assume that Vf ⊂ Vλ for some
eigenvalue λ of 4M . Then f : M → V is an isometric immersion with respect
to c times the original metric on M . By Takahashi’s theorem [20] f maps into a
sphere rSV for some r . Calculating 4M (|f |2 ), we obtain c = r2 λ/ dim M . If f is
normalized then r = 1 and we get c = λ/ dim M . Again by Takahashi, we obtain
that f : M → SV is an isometric minimal immersion of the λ/ dim M -multiple of
the metric on M .
Let H ⊂ Vλ be a δ -conformal G-submodule. By definition, δH : M → H is
conformal with VδH = H ⊂ Vλ so that the argument above applies. Since δH is
automatically normalized, we obtain that δH : M → SH is an isometric minimal
immersion of the λ/ dim M -multiple of the metric on M . In particular, H is
δ -spherical.
Remark.
Let M = G/K be isotropy irreducible. Then any irreducible
GPN
∞
j
submodule H ⊂ C (M ) is δ -conformal. Indeed, (12) holds because j=0 dχ
dχj is a G-invariant bilinear form on H. Its coordinate representation in (5) shows
that δH : M → SH∗ is the standard minimal immersion [6,25].
A DoCarmo-Wallach type argument gives the following:
Toth
563
Proposition 3.4.
Let H ⊂ Vλ ⊂ C ∞ (M ) be a δ -conformal G-submodule.
Then the congruence classes of isometric minimal H-immersions f : M → SV
(with respect to the λ/ dim M -multiple of the metric on M ) are parametrized by
the compact convex body
M(H) = K0 (H) ∩ F(H),
(13)
in the G-module
F(H) = {XδH
Y δH | X, Y ∈ T (M )}⊥ ⊂ S 2 (H).
(14)
F(H) ⊂ E(H),
(15)
We also have
so that
M(H) = L(H) ∩ F(H).
Let M = G/K be a naturally reductive homogeneous space with orthogonal
decomposition g = k ⊕ p, where g and k are the Lie algebras of G and K , and
p is identified with the tangent space To (M ), o = {K}. The subgroup K acts
on its Lie algebra k by the adjoint representation, and, under the identification
p∼
= To (M ), this action corresponds to the action of K on To (M ) via the isotropy
representation.
As noted in Section 1, if W is any (finite dimensional) orthogonal K -module then
the induced G-module IndG
K (W) is comprised of continuous maps f : G → W that
satisfy f (kg) = k · f (g), g ∈ G, k ∈ K . Precomposition of these maps by right
multiplication on G defines the G-module structure on IndG
K (W). In addition,
integration with respect to the Haar measure on G and the scalar product on W
define a G-invariant scalar product on IndG
K (W).
By Frobenius reciprocity, we have
HomG (V, IndG
K (W)) = HomK (V|K , W),
where V is an orthogonal G-module and W is an orthogonal K -module.
Let H ⊂ C ∞ (M ) be a δ -conformal G-module. Restricting the differential of δH
to p = To (M ) gives a K -equivariant linear imbedding (δH )∗ : p → H. We identify
the K -module p with the image, and think of p as a K -submodule of H|K . Notice
that this can also be thought of as the inclusion p ⊂ H∗ ∼
= H given by the action of
the tangent vectors at o to M on the elements of H by directional differentiation.
We define
2
Ψ : S 2 (H) → IndG
K (S (p))
(16)
as follows. For C ∈ S 2 (H), we let Ψ(C) : G → S 2 (p) be the map defined by
Ψ(C)(g) = π(g · C), where π : S 2 (H) → S 2 (p) is the orthogonal projection, a
homomorphism of K -modules.
We have
ker Ψ = F(H).
564
Toth
Indeed, for C ∈ S 2 (H), Ψ(C) = 0 if and only if hg · C, S 2 (p)i = 0 for all g ∈ G, if
and only if hC, g · S 2 (p)i = 0 for all g ∈ G. In view of the identification p ⊂ H|K ,
this holds if and only if
hC, S 2 ((δH )∗ (Tx (M )))i = 0,
x ∈ M.
This is equivalent to C ∈ F(H).
4.
Proofs of Theorems A-C.
Proof of Theorem A. We first let (G, K) = (SO(m + 1), S(m)), SO(m) =
SO(m) ⊕ [1] ⊂ SO(m + 1), with M = G/K = S m the Euclidean m-sphere, and
H = Vλp . The eigenspace Vλp corresponding to λp = p(p + m − 1) is Hp , the
irreducible SO(m + 1)-module of spherical harmonics of order p on S m .
We let P p denote the SO(m + 1)-module of homogeneous polynomials on Rm+1
of degree p (with the usual action g · ξ = ξ ◦ g −1 , g ∈ SO(m + 1), ξ ∈ P p ).
By homogeneity, a polynomial in P p is uniquely determined by its restriction to
S m ⊂ Rm+1 .
We also think of a spherical harmonic χ of order p on S m as a harmonic homogeneous polynomial on Rm+1 of degree p. (The equivalence of these two representations is given by restriction from Rm+1 to S m , and comparison of the Euclidean
and spherical Laplacians. We suppress the restriction if there is no danger of confusion.) This way Hp becomes an SO(m + 1)-submodule of P p . We have the
orthogonal decomposition
[p/2]
p
p
P =H ⊕P
p−2
2
·ρ =
X
Hp−2k · ρ2k ,
(17)
k=0
where ρ(x) = |x|, x = (x0 , . . . , xm ) ∈ Rm+1 [15,24].
Since Hp ⊂ P p , we have
p
p
H ·H ⊂P
2p
=
p
X
H2j
j=0
as SO(m + 1)-modules. Theorem A for M = S m states that equality holds, and
this is what we need to show.
We define the harmonic projection operator as the orthogonal projection H : P p →
Hp with kernel ker H = P p−2 · ρ2 [24]. It is given explicitly by
[p/2]
H(ξ) = ξ +
X (−1)j (p − 1) . . . (p − j)
j=1
j!λ2(p−1) . . . λ2(p−j)
4j ξ · ρ2j ,
ξ ∈ P p.
(18)
Since SO(m) fixes xm , a zonal spherical harmonic in Hp is H(xpm ). By (18), it is
given by
H(xpm ) = xpm
[p/2]
+
X (−1)j (p − 1) . . . (p − j)p(p − 1) . . . (p − 2j + 1)
j=1
j!λ2(p−1) . . . λ2(p−j)
xp−2j
ρ2j .
m
565
Toth
Rewriting the coefficients in terms of the Gamma function, we obtain
[p/2]
X (−1)j Γ p + m−1 − j
p!
p
2
H(xm ) = p
(2xm )p−2j ρ2j .
m−1
j!(p
−
2j)!
2 Γ p+ 2
j=0
Up to a normalizing factor, this is the ultraspherical polynomial Cpν with ν =
(m − 1)/2 [1,24]:
p!Γ m−1
p
2
ρp Cp(m−1)/2 (cos θ),
H(xm ) = p
(19)
m−1
2 Γ p+ 2
where xm /ρ = cos θ . In terms of the Jacobi polynomials, we have
Cpν =
(2ν)p
P (ν−1/2,ν−1/2) ,
(ν + 1/2)p p
(20)
where (a)p = Γ(a + p)/Γ(a). The choice of the zonal spherical harmonic χ0 for
M = S m specified in the first table of Section 2 follows. The linearization of the
product formula for ultraspherical polynomials [7] reads as
min (p,q)
Cpν Cqν
=
X (p + q + ν − 2k)
(p + q + ν − k)
k=0
×
(ν)k (ν)p−k (ν)q−k (2ν)p+q−k (p + q − 2k)! ν
C
.
k!(p − k)!(q − k)!(ν)p+q−k (2ν)p+q−2k p+q−2k
We now let p = q and ν = (m − 1)/2. In view of (20), the linearization
formula above reduces to (2) with α = β = m/2 − 1, and we also obtain an
explicit formula for the linearization coefficients. This immediately shows that
c(j, p; m/2 − 1, m/2 − 1) is nonzero if and only if 0 ≤ j ≤ 2p is even. By (19),
evaluating (2) on cos θ , the Jacobi polynomials become zonal spherical harmonics.
(m/2−1,m/2−1) 2
Suppressing the argument cos θ , by definition, (Pp
) ∈ Hp · Hp . The
2p
2j
restriction of the orthogonal projection P → H , j = 0, . . . , p, to Hp ·Hp ⊂ P 2p
(m/2−1,m/2−1) 2
(m/2−1,m/2−1)
maps (Pp
) to a nonzero constant multiple of P2j
since
2j
c(2j, p; m/2 − 1, m/2 − 1) is nonzero. Schur’s lemma implies that H must be a
component of Hp · Hp for j = 0, . . . , p. Theorem A follows for M = S m .
For M = RP m , the real projective m-space, the eigenspace Vλp corresponding to
m
the p-th eigenvalue λp = 2p(2p + m − 1) of the Laplacian 4RP can be identified
with H2p . Theorem A follows from the spherical case above.
Next we let (G, K) = (U (m + 1), U (m) × U (1)) with CP m = U (m + 1)/(U (m) ×
U (1)), the complex projective m-space. Let P p,q denote the space of complex
homogeneous polynomials of bidegree (p, q) on Cm+1 . An element ξ ∈ P p,q is
a complex valued homogeneous polynomial that has degree p in the variables
z0 , . . . , zm ∈ C and degree q in the variables z̄0 , . . . , z̄m ∈ C. By homogeneity, ξ
can be thought of as a function on the unit sphere S 2m+1 ⊂ Cm+1 .
The space P p,p is the complexification of a real U (m + 1)-submodule, and this
real submodule is also denoted by the same symbol. An element in P p,p can be
thought of as a function on CP m .
The decomposition in (17) gives
min (p,q)
P
p,q
=H
p,q
⊕P
p−1,q−1
2
·ρ =
X
k=0
Hp−k,q−k · ρ2k ,
566
Toth
where ρ = |z|, z = (z0 , . . . , zm ) ∈ Cm+1 , and Hp,q is the space of complex
harmonic homogeneous polynomials of bidegree (p, q) on Cm+1 . Then Hp,q is a
complex irreducible U (m + 1)-module. For real valued polynomials we also have
P
p,p
=
p
X
Hp−k,p−k · ρ2k ,
k=0
as real U (m+1)-modules. Here P p,p is the space of real valued homogeneous polynomials of bidegree (p, q) on Cm+1 , and Hj,j is the eigenspace Vλj corresponding
to the j -th eigenvalue λj = 4j(j + m).
Since Hp,p ⊂ P p,p , we have
H
p,p
·H
p,p
⊂P
2p,2p
=
2p
X
Hj,j
(21)
j=0
as real U (m + 1)-modules. To prove Theorem A for M = CP m , it remains to
show that equality holds.
Since U (m) fixes zm and the center U (1) acts on Hp,p trivially, a zonal spherical
harmonic in Hp,p is H(|zm |2p ). Here the harmonic projection operator H is the
restriction of the harmonic projection for the spherical case above. We have
H(|zm |2p ) =
p!(p + m − 1)! 2p (m−1,0)
ρ Pp
(cos(2θ)),
(2p + m − 1)!
where |zm |/ρ = cos θ . (See also [24], formula (5’) in Chapter 11.3.2, Vol.2.) In the
(m−1,0) 2
linearization of the square (Pp
) all coefficients c(j, p; m−1, 0), j = 0, . . . , 2p,
are positive for m ≥ 2. As in the spherical case it follows that Hj,j , j = 0, . . . , 2p,
are U (m + 1)-submodules of Hp,p · Hp,p . The equality in (21) follows.
The cases of the quaternionic projective space HP m and the Cayley projective
plane CaP 2 are entirely analogous [10]. The zonal spherical functions for HP m
are explicitly derived in [24] (cf. formula (14) in Chapter 11.7.4, Vol. 2). Another
approach for the Cayley projective plane is to determine the highest weights of
the class 1 modules with respect to the pair (F4 , Spin(9)) and to use the Weyl
dimension formula for the multiplicities.
Proof of Theorem B. One of the principal results of [22] (Theorem 4.1, p.
136) gives the multiplicity m of Hq,q in S 2 (Hp,p ) as follows:
q,q 2 p,p 1
1 + (−1)q
m H : S (H ) =
min (q, 2p − q) + 1 +
.
2
2
By the definition of Ē(Hp,p ), we thus have
2
p,p
S (H ) =
2p
X
1
q=0
2
1 + (−1)q
min (q, 2p − q) + 1 +
2
Hq,q ⊕ Ē(Hp,p ).
On the other hand, by Proposition 3.3 and Theorem A, we have
!
2p
X
Vλq .
E(Vλp ) = S 2 (Vλp )/(Vλp · Vλp ) = S 2 (Vλp )/
q=0
567
Toth
Since Vλq = Hq,q , combining these two formulas, Theorem B follows.
Proof of Theorem C. The proof is based on comparing the multiplicities of
some irreducible components of the domain and the image of Ψ in (16) . To do
this, we first complexify, and consider
U (m+1)
Ψ : S 2 (Hp,p ) ⊗R C → IndU (m)×U (1) (S 2 (Cm ) ⊗R C),
(22)
where Cm = To (CP m ) with U (m) × U (1)-module structure given by U (m) acting
on Cm by matrix multiplication, and the center U (1) ⊂ U (m + 1) acting trivially.
For the irreducible decompositions, recall that a complex irreducible U (m + 1)module V is given by its highest weight which, with respect to the standard
maximal torus in U (m + 1), is an (m + 1)-tuple with integral coefficients. We
write V ρ = VUρ(m+1) , where ρ = (ρ1 , . . . , ρm+1 ) ∈ Zm+1 with
ρ1 ≥ ρ2 ≥ . . . ≥ ρm+1 .
P
The center U (1) ⊂ U (m + 1) acts by the weight m+1
j=1 ρj .
For example, we have
Hp,q = V (p,0,... ,0,−q) .
The branching rule for restrictions from U (m + 1) to U (m) takes the form
X
VUρ(m+1) |U (m) =
VUσ(m) ,
σ
where the summation runs over all σ ∈ Zm for which
ρ1 ≥ σ1 ≥ ρ2 ≥ . . . ≥ ρm ≥ σm ≥ ρm+1 .
The decomposition of the domain in (22) into irreducible components is one of the
technical results in [22] (Theorem 4.1 on p. 136). For m ≥ 3, we have
S (H ) ⊗R C ∼
=
2
p,p
2p min (b,2p−b) min (b,p,e)
X
X
X n0 (b, c, d) + m0 (b, c, d)
c=0
b=0
d=0
(b,c,0,... ,0,−d,d−b−c)
× V
Here e =
b+c 2
2
(23)
.
, and
n0 (b, c, d) = min (b − c, b − d, p − c, p − d, b + c − 2d, 2p − b − c) + 1.
m0 (b, c, d) = 0 for b 6≡ c( mod 2), and for b ≡ c( mod 2), we have
−1 if b, d are odd and m ≡ 1( mod 4)
m0 (b, c, d) =
1
otherwise.
We now fix a component V (b,c,0,... ,0,−d,d−b−c) in S 2 (Hp,p ) ⊗R C. We need to
determine the multiplicity
h
i
U (m+1)
m V (b,c,0,... ,0,−d,d−b−c) : IndU (m)×U (1) (S 2 (Cm ) ⊗R C) .
(24)
568
Toth
First note that the multiplicity in (24) is the dimension of the module
HomU (m) V (b,c,0,... ,0,−d,d−b−c) |U (m) , S 2 (Cm ) ⊗R C .
(25)
This follows by Frobenius reciprocity along with the fact that U (1) acts trivially.
In particular, the multiplicity in (24) is nonzero if and only if V (b,c,0,... ,0,−d,d−b−c)
is disjoint from F̄(Hp,p ).
As a real SO(2m)-module
S 2 (Cm ) = S 2 (R2m ) = H0 ⊕ H2 .
Complexifying, and restricting to U (m) ⊂ SO(2m), we obtain
2
m
0
2
S (C ) ⊗R C = H |U (m) ⊕ H |U (m) = H
0,0
⊕
2
X
H2−j,j .
j=0
Thus (25) can be written as
HomU (m)
V (b,c,0,... ,0,−d,d−b−c) |U (m) , H0,0 ⊕
2
X
H2−j,j
!
.
j=0
The dimension of this module is equal to
2
X
m H0,0 : V (b,c,0,... ,0,−d,d−b−c) |U (m) +
m H2−j,j : V (b,c,0,... ,0,−d,d−b−c) |U (m) .
j=0
By the branching rule, the first multiplicity is 1 if and only if c = d = 0 and
zero otherwise. The remaining multiplicities can be obtained similarly using the
branching rule. For 0 ≤ j ≤ 2, we obtain
2−j,j
1 if b ≥ 2 − j ≥ c and −d ≥ −j ≥ d − b − c
(b,c,0,... ,0,−d,d−b−c)
m H
:V
|U (m) =
0 otherwise.
Comparing this with (23), we see that, for m 6≡ 1 (mod 4), p = 3, b = 3,
c = d = 1, the multiplicity of the component V (3,1,0,... ,0,−1,−3) is 2 in the domain
of Ψ and 1 in the image of Ψ. The same holds for p = 4, b = 4, c = d = 1.
Theorem C follows.
Acknowledgment. The author wishes to thank Professor Gasper for pointing
out positivity of the linearizing coefficients in products of Jacobi polynomials. The
author is indebted to Professor Knop for various discussions on representation
theory, and for providing the multiplicity formulas for the eigenvalues in the
quaternionic and Cayley cases. The author is also indebted to the referee for
the careful reading of the manuscript and for pointing out various improvements
of the text.
References
[1]
Askey, R., “Orthogonal Polynomials and Special Functions,” CBMS-NSF
Reg. Conf. Ser. in Applied Math. 21 SIAM, 1975.
Toth
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
569
Besse, A., “Manifolds all of whose Geodesics are Closed,” Springer-Verlag,
Berlin etc., 1978.
Calabi, E., Minimal immersions of surfaces in euclidean spheres, J. Diff.
Geom. 1 (1967), 111–125.
Cartan, É., Sur la détermination d’une système orthogonal complet dans
un espace de Riemann symétrique clos, Rend. Circ. Mat. Palermo 53
(1929), 217–252.
Chavel, I., “Eigenvalues in Riemannian Geometry,” Academic Press, New
York etc. 1984.
DoCarmo, M., and N. Wallach, Minimal immersions of spheres into spheres,
Ann. of Math. 93 (1971), 43–62.
Dougall, J., A theorem of Sonine in Bessel functions, with two extensions
to spherical harmonics, Proc. Edinburgh Math. Soc. 37 (1919), 33–47.
Eells, J., and J. H. Sampson, Harmonic mappings of Riemannian manifolds, Amer. J. Math. 86 (1964), 109–160.
Escher, Ch., and G. Weingart, Orbits of SU (2)-representations and minimal isometric immersions, Math. Ann. 316 (2000), 743–769.
Gangolli, R., Positive definite kernels on homogeneous spaces and certain
stochastic processes related to Lévi’s Brownian motion on several parameters, Ann. Inst. H. Poincaré, Sect. B 3 (1967), 121–226.
Gasper, G., Linearization of the product of Jacobi polynomials I, Can. J.
Math. 22 (1970), 171–175.
Gasper, G., Linearization of the product of Jacobi polynomials II, Can. J.
Math. 22 (1970), 582–593.
Gasper, G., private communication.
Gauchman, H., and G. Toth, Fine structure of the space of spherical
minimal immersions, Trans. Amer. Math. Soc. 348 (1996), 2441–2463.
Helgason, S., “Groups and Geometric Analysis. Integral Geometry, Invariant Differential Operators, and Spherical Functions,” Mathematical
Surveys and Monographs 83 American Mathematical Society, Providence,
2000.
Helgason, S., “Geometric Analysis on Symmetric Spaces,” Mathematical
Surveys and Monographs 39 American Mathematical Society, Providence,
1994.
Helgason, S., The Radon transform on Euclidean spaces, compact two-point
homogeneous spaces and Grassmann manifolds, Acta Math. 113 (1965),
153–180.
Mashimo, K., Degree of a standard isometric minimal immersions of the
symmetric spaces of rank one into spheres, Tsukuba J. Math. Vol. 5 (1981),
291–297.
Rahman, M., A non-negative representation of the linearization coefficients
of the product of Jacobi polynomials, Can. J. Math. 33 (1981), 915–928.
Takahashi, T., Minimal immersions of Riemannian manifolds, J. Math.
Soc. Japan 18 (1966), 380–385.
570
[21]
[22]
[23]
[24]
[25]
[26]
Toth
Toth, G., Eigenmaps and the space of minimal immersions between spheres,
Indiana Univ. Math. J. 46 (1997), 637–658.
Toth, G., “Harmonic Maps and Minimal Immersions through Representation Theory,” Academic Press, New York etc., 1990.
Toth, G., and W. Ziller, Spherical minimal immersions of the 3-sphere,
Comment. Math. Helv. 74 (1999), 1–34.
Vilenkin, N.I., and A. U. Klimyk, “Representation of Lie Groups and
Special Functions,” Vols. 1-3, Kluwer, Dordrecht, 1991.
Wallach, N., Minimal immersions of symmetric spaces into spheres, in
“Symmetric Spaces”, Marcel Dekker, New York (1972), 1–40.
Weingart, G., “Geometrie der Modulräume minimaler isometrischer Immersionen der Sphären in Sphären,” Bonner Mathematische Schriften 314,
Bonn, 1999.
Gabor Toth
Department of Mathematical Sciences
Rutgers University
Camden, New Jersey, 08102
gtoth@camden.rutgers.edu
Received August 3, 2001
and in final form November 29, 2001
Download